Domino Data Lab: Enterprise AI at scale...in one click

Human progress is fueled by innovation, with steady advancements across millennia punctuated by spikes of inventions that re-shape the future overnight—big ideas with big impact like the printing press, the steam engine, the Internet. Many believe the industrialization of artificial intelligence (AI) fueling data science is one of those milestones. Which is why many companies now require employees to use AI tools first when hiring new talent, seeking clear answers to complex questions, formalizing go-to-market strategies, and designing next-generation products.

There is exciting news on that front. Domino Data Lab (Domino) and NetApp saw something deeper, something with broader potential beyond the froth of ChatGPT everywhere and ubiquitous AI washing.​ ​They recognized that one of the most significant challenges to moving AI workloads from prototype into production was managing and accessing data efficiently, securely, and seamlessly across diverse environments.​​

​​​​Most enterprises today operate with data distributed across multiple environments, and efficient data access remains a persistent challenge. The Domino and NetApp partnership uniquely addresses this critical issue by enabling AI teams to seamlessly access governed datasets, simplify infrastructure management, and accelerate innovation at scale—all without compromising security or compliance from IT’s perspective.​​​ With AI well along in its hype cycle, a new phase of AI adoption is underway, with businesses, governments, and industries seeking partners ​​to help them move AI prototypes into production at scale.

The partnership has resulted in the first integration of Domino’s leading MLOps platform with NetApp’s advanced underlying data management solution. Data scientists gain self-serve capabilities to manage AI workloads directly within Domino using Domino Volumes for NetApp^® ONTAP^® (DVNO). ​​Domino uses Amazon FSx for NetApp^® ONTAP^® as the foundation for its SaaS offering, ​​resulting in a 2x performance improvement compared​​ to alternatives, while preserving full traceability of their models and agents. This streamlined approach allows data scientists and researchers to focus on innovation rather than infrastructure.​​​

This milestone in AI, machine learning operations (MLOps), and data storage represents a significant advancement. It stems from the recognition that unlocking the full promise of AI requires seamlessly integrated capabilities: robust compute power, efficient data management, and streamlined workflows that give data scientists self-service freedom while IT maintains operational integrity using standard, trusted practices – reducing data sprawl and accelerating AI deployment. ​​​

It all begins when peers at NVIDIA recommend friends at both Domino and NetApp have lunch. In short order, leaders at the companies realize​d​ they share​d​ a vision for a future in which AI compute and AI storage are, by definition, hybrid or multi-cloud. Their vision promis​​ed​​​ a new way for data scientists and storage engineers to jointly usher in a new era of data science.

Domino, the leading enterprise MLOps platform, has the ​​AI lifecycle management and compute orchestration​​​ ​expertise. And NetApp, the intelligent data infrastructure company, has the data storage expertise. But the partnership is more than a meeting of minds, it’s also a match of technical capabilities. Thomas ‘T-Rob’ Robinson, chief operating officer for Domino describes the partnership as akin to, “a peanut butter and jelly sandwich (or, for the international set, Marmite and cheese).” It’s simply good.

This vision for a new way forward means data scientists and organizations today can easily and confidently move beyond prototype models to autonomously functioning agents in the real world, where AI on a unified platform can make a real difference.

Thomas Been, chief marketing officer for Domino says, “One challenge hindering successful AI adoption by many organizations is trust.” Trust that AI results discovered at the pilot level will scale to the same results at the agent level, where regulatory governance and hard costs are especially critical considerations.

Just as challenging, most AI pilots are run in the cloud, an environment proven for data scientists to be agile, scalable, and cost-effective for elastic pay-as-you-need processing and storage. The cloud also enables disparate teams to access and share data in fluid, highly collaborative ways.

As a result, businesses and organizations often get tantalizing glimpses of AI’s value when standing up prototypes in the cloud. But, run in isolation and using carefully curated data sets, models are prone to over-training and learned biases, to name just two pitfalls. Unlocking the value at scale outside of the test environment remains more elusive, as larger datasets introduce variations a model doesn’t recognize. Model hallucinations, inaccurate predictions, and diminished effectiveness explain why 85% of AI investments remain stuck in pilot mode. Thomas notes, “Most organizations are not considering all of their options by taking their data to AI.” Out of habit, that means to the cloud. But with Amazon FSx for NetApp^® ONTAP^®, old habits are changing.

They’re doing this in hybrid and multi-cloud environments, because most of the world’s data that matters the most is still on-premises. As Jensen Huang, CEO of NVIDIA notes, “Nearly half of the files in the world are stored on-prem on NetApp.” This is why Domino and NetApp’s joint customers are taking a different path after considering their options. Thomas says, “They’re bringing AI to their data.” One example is the way in which Domino's AI and data science platform scales in hybrid and multi-cloud environments where the most intensive enterprise AI workloads require high throughput, especially when data scientists are running multiple models against common datasets spanning on-premises and the cloud. Amazon FSx for NetApp^® ONTAP^® enables feeding the models as fast as GPUs process the data, a game changer on its own, but there’s even more to the story.

The currency cost of AI computation varies widely, from entry-level GPUs to high-end options, both for the actual price tag of a chip, and for time needed to process a dataset, whether in the cloud or onsite. But Thomas raises a topic of equal importance. He says, “You’ve also got to add in the cost of people time.”

In the past, data scientists followed a highly linear process, beginning with adherence to regulatory governance, company policy, and security protocols, and including managed access to data lakes and warehouses ​​through approved tools like SQL and Tableau. This process of requesting, gatekeeping, waiting, and downloading can add hours, days, even weeks to a data scientist’s research project. Not only can these roadblocks spur data scientists to seek out shadow IT, they also significantly slow down time to production. Finally, every move of data from one environment to another introduces risks to the data itself, from data loss and duplication to corruption through exposure to systems that are less secure. Now consider that IDC finds the typical data scientist goes through this process 7-10 times in building a single model. The numbers—from TBs and GPUs to days and dollars—can be significant, especially when multiplied by an organization’s number of data scientists and their numerous current experiments.

NetApp developed ‘snapshotting ‘as a core feature of the ONTAP operating system years ago. That is, they empowered data scientists to make instant copies of their data at any point in time without the hassle of futzy and time-consuming ‘save as’ processes. Today saving these immutable copies is a central value proposition of the Domino solution through Amazon FSx for NetApp^® ONTAP^®—zero footprint snapshots of data and data models captured on the fly in just one click. This is invaluable not only for audit and reference purposes, but also for reverting to earlier, more useful models, and for sharing with colleagues. Now data scientists can do research, and storage engineers can focus on managing and securing their domain.

The FSx for NetApp^® ONTAP^® solution enables AI at scale in a repeatable MLOps lifecycle. It also takes advantage of the wealth of intelligent data management technology available in NetApp without users having to know anything about storage or how to administer volumes. A pharma customer of Domino’s said, “This is exactly what we need!” FSx for NetApp^® ONTAP^® is the result of a product-driven strategy to empower users to do more for less—that is, work faster, and finish faster, with lower costs. It’s the opposite of what came before. Namely, generic hyperscaler storage with lower performance, GPU bottlenecks, and inefficiencies. Re-platforming with the Domino and NetApp solution already makes a difference for early adopters who now realize the benefits of models run at scale against their hybrid domain, all with faster throughput and no wasted GPU capacity.

And lost in the headlines of AI’s surging energy consumption, it turns out AI can be greener than its reputation, because higher throughput and faster processing also reduce an organization’s environmental footprint, not just their GPU and power expenses. FSx for NetApp^® ONTAP^® can be one more step for a company to take in its sustainability journey.

Thomas Been sees 2025 as an inflection point with a ‘before’ and ‘after’, when organizations move beyond the ‘artisanal era’ of AI and Industrial AI commences. He says, “We’re excited about our partnership with NetApp, because we changed the game, together. Domino and NetApp have created standardized processes—with the right governance throughout—to enable enterprises to embrace AI at scale.” But Domino’s offering through Amazon FSx for NetApp^® ONTAP^® is just the beginning. What’s next?

First up is product design work to integrate Domino’s capabilities directly within NetApp ONTAP, as well as a suite of NetApp-aware features for MLOps pros to use as they further mature their pilot-to-agent AI pipeline.

While it may not be the Singularity, FSx for NetApp^® ONTAP^® is a singular solution bringing unique features and valuable benefits to data scientists looking to increase crop yields, design safer drugs, and accelerate production of autonomous vehicles. T-Rob sums things up, saying, “The immense value of a stack with Domino and NetApp, and partners like AWS and NVIDIA, comes from it being an enterprise foundation with all the resilience, efficiency, and security that AI needs.” Peanut butter and jelly, indeed.